wisio 0.1.1

Analyze, visualize, and understand I/O performance issues in HPC workflows

WisIO: Workflow I/O Analysis Tool

Overview

WisIO (Wisdom from I/O Behavior) is an open-source tool designed to efficiently analyze multi-terabyte-scale workflow performance data over distributed resources. It provides a comprehensive analysis of I/O performance, identifying bottlenecks and potential root causes through advanced rule-based analysis. With its extensible design, WisIO can be tailored to various use cases, providing actionable insights for improving application performance and resource utilization. By leveraging parallel computing and multi-perspective views, WisIO enables rapid detection of complex I/O issues, making it an invaluable asset for HPC professionals and researchers.

Installation

To install WisIO through pip (recommended for most users):

# Ensure runtime dependencies for optional features (e.g., Darshan, Recorder) are installed.
# This might involve using your system's package manager or a tool like Spack.
# Example using Spack to prepare the environment:
# spack -e tools install
pip install wisio[darshan,dftracer]

To install WisIO from source (for developers or custom builds):

# 1. Install system dependencies:
#    Refer to the "Install system dependencies" step in .github/workflows/ci.yml
#    (e.g., build-essential, cmake, libarrow-dev, libhdf5-dev, etc.).
#    Alternatively, tools like Spack can help manage these:
#    # spack -e tools install

# 2. Install Python build dependencies:
python -m pip install --upgrade pip meson-python setuptools wheel

# 3. Install WisIO from the root of this repository:
#    The following command includes optional C++ components (tests and tools).
#    The --prefix argument is optional and specifies the installation location.
pip install .[darshan,dftracer] \
  -Csetup-args="--prefix=$HOME/.local" \
  -Csetup-args="-Denable_tests=true" \
  -Csetup-args="-Denable_tools=true"

# (Optional) Install dependencies for running tests if you plan to contribute or run local tests:
# pip install -r tests/requirements.txt

Usage

Here's an example of how to run WisIO with the recorder analyzer using sample data included in the repository:

# Before running, ensure the sample data is extracted.
# For example, to extract the 'recorder-parquet' sample used below:
# mkdir -p tests/data/extracted 
# tar -xzf tests/data/recorder-parquet.tar.gz -C tests/data/extracted
wisio +analyzer=recorder percentile=0.99 trace_path=tests/data/extracted/recorder-parquet

This command will analyze the traces and print a summary of I/O characteristics and detected bottlenecks. Below is a sample of the "I/O Characteristics" output:

╭───────────────────────────────────── CM1 I/O Characteristics ─────────────────────────────────────╮
│                                                                                                   │
│  Runtime          667.81 seconds                                                                  │
│  I/O Time         4.12 seconds                                                                    │
│                   ├── Read - 0.00 seconds (0.05%)                                                 │
│                   ├── Write - 0.58 seconds (14.08%)                                               │
│                   └── Metadata - 3.53 seconds (85.89%)                                            │
│  I/O Operations   27,463 ops                                                                      │
│                   ├── Read - 1,282 ops (4.67%)                                                    │
│                   ├── Write - 2,303 ops (8.39%)                                                   │
│                   └── Metadata - 23,878 ops (86.95%)                                              │
│  I/O Size         21.18 GiB                                                                       │
│                   ├── Read - 20.03 GiB (94.59%)                                                   │
│                   └── Write - 1.15 GiB (5.41%)                                                    │
│  Read Requests    4 MiB-16 MiB - 1,282 ops                                                        │
│                   └── 4-16 MiB - 1,282 ops (100.00%)                                              │
│  Write Requests   4 kiB-16 MiB - 2,303 ops                                                        │
│                   ├── <4 kiB - 397 ops (17.24%)                                                   │
│                   ├── 4-16 kiB - 1,092 ops (47.42%)                                               │
│                   ├── 16-64 kiB - 722 ops (31.35%)                                                │
│                   ├── 64-256 kiB - 1 ops (0.04%)                                                  │
│                   └── 4-16 MiB - 91 ops (3.95%)                                                   │
│  Nodes            1 node                                                                          │
│  Apps             1 app                                                                           │
│  Processes/Ranks  1,280 processes                                                                 │
│  Files            775 files                                                                       │
│                   ├── Shared: 38 files (4.90%)                                                    │
│                   └── FPP: 737 files (95.10%)                                                     │
│  Time Periods     393 time periods (Time Granularity: 10,000,000.0)                               │
│  Access Pattern   Sequential: 3,585 ops (100.00%) - Random: 0 ops (0.00%)                         │
│                                                                                                   │
╰─ R: Read - W: Write - M: Metadata  ───────────────────────────────────────────────────────────────╯

WisIO also identifies potential I/O bottlenecks. Here is a snippet of the "I/O Bottlenecks" section from the same run:

╭────────────────── I/O Operations per Second: 25 I/O Bottlenecks with 56 Reasons ──────────────────╮
│                                                                                                   │
│  Time View (4 bottlenecks with 7 reasons)                                                         │
│  ├── [CR1] 32 processes access 2 files within 1 time period (5) across 32 I/O operations and      │
│  │   have an I/O time of 2.19 seconds which is 53.26% of overall I/O time of the workload.        │
│  │   └── [Excessive metadata access] Overall 100.00% (2.19 seconds) of I/O time is spent on       │
│  │       metadata access, specifically 100.00% (2.19 seconds) on the 'open' operation.            │
│  ├── [CR2] 1 process accesses 6 files within 1 time period (634) across 40 I/O operations and     │
│  │   has an I/O time of 0.33 seconds which is 7.97% of overall I/O time of the workload.          │
│  │   ├── [Excessive metadata access] Overall 99.35% (0.33 seconds) of I/O time is spent on        │
│  │   │   metadata access, specifically 99.13% (0.33 seconds) on the 'open' operation.             │
# ... (further bottleneck details omitted for brevity) ...
│                                                                                                   │
╰───────────────────────────────────────────────────────────────────────────────────────────────────╯

Further Information

For more details, to report issues, or to contribute to WisIO, please refer to the following resources:

• Official WisIO Documentation: For detailed usage, configuration options, and information about analyzers.
• Issue Tracker: To report bugs or suggest new features.
• Contributing Guidelines: For information on how to contribute to the project, including setting up a development environment and coding standards.
• Citation File: If you use WisIO in your research, please cite it using the information in this file.

Acknowledgments

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research under the DOE Early Career Research Program (LLNL-CONF-862440). Also, this research is supported in part by the National Science Foundation (NSF) under Grants OAC-2104013, OAC-2313154, and OAC-2411318.